1. Field of the Invention
The present invention relates to a process for converting linear coded pulse code modulated (PCM) words into non-linear coded PCM words (companding) and vice versa for converting non-linear coded PCM words into linear coded PCM words (linearization) in accordance with a multiple segment curve, employing a shift register into which, in the event of companding, a linear coded PCM word is input and is shifted therein until a significant bit appears at the series output of the shift register, whereupon, in order to form the non-linear coded PCM words, those bit combinations which characterize the number of shift clock pulses required for this purpose are used as higher-value word components, and a number of bits which follow the aforementioned significant bit and which are contained in the shift register are used as lower-value word components, and in the event of linearization, the lower-value word component, which corresponds to the grading within the segments of the aforementioned curve, of a non-linear coded PCM word is input and is shifted therein by a number of positions which is linked to the particular segment number of the curve designated by that part of the non-linear coded PCM word which has not been taken into consideration, whereupon, with a value changed in this manner, it is used to form the higher-value component of the linear coded PCM word.
2. Description of the Prior Art
The type of conversion mentioned above of linear code representations into non-linear code representations and vice-versa is of particular interest in the context of the digital switching of conference conversations in the course of which sum code words must be formed from the PCM words which are supplied by the individual conference subscribers and which, in the interest of constant signal-to-noise ratio, must possess non-linear code representation. However, this sum formation produced conclusive results only when the individual PCM words which are to be added have previously been subjected to linearization. The further transmission of the sum code word must then be preceded by a reconversion into non-linear code representation.
In terms of storage expense required for its implementation, the above mentioned process (German Offenlegungsschrift No. 1,803,222 corresponding to U.S. Pat. No. 3,575,591 and fully incorporated herein by this reference) has advantages in comparison to a process based on storing the assignment of non-linear coded PCM words to linear coded PCM words even if two proposed process variants (German patent application No. P 29 38 984.2 corresponding to U.S. Ser. No. 178,531 filed Aug. 15, 1980 and the German patent application No. P 30 28 726.4) are adopted which lead, in essence, to a reduction in the storage space requirement in comparison to the normal situation in which the assignment to a non-linear coded word is stored for each linear coded value.
The conversion procedure referred to above is based on a 13-segment curve which obeys the so-called A-law and which, because of its regularity, is particularly suitable for such conversion since, apart from the first segment in one-half of the curve, the conversions can be carried out in both directions in that at least one part of the PCM word which is to be converted is input into a shift register and, depending upon the direction of conversion, is shifted therein in dependence upon the segment upon which the value represented by the PCM word in question is located, or else is shifted until a bit possessing a specific value occurs at the shift register output, whereupon the bits which have been shifted in this manner, in their modified values and possibly together with a bit combination corresponding to the number of shifts, form the recoded PCM word. As indicated above, this procedure is not possible in the case of PCM words which are assigned to the first segment of one-half of the curve. If such PCM words are to be subjected to conversion, instead, depending upon the direction of the conversion, further specified bit combinations are attached to the bits of the non-linear coded PCM words or individual bit values within a linear coded PCM word which is to be converted are suppressed.
In the PCM time division multiplex (TDM) systems which have been introduced in various countries, the digital/analog conversion is based on the so-called .mu.-law which differs from the above-mentioned A-law in that the relationship between non-linear coded values and coded values is not governed by 13-segment curve but by a 15-segment curve. In this case, the 128 non-linear coded values are opposed by twice as many linear coded values as in the case of the A-law, namely 8159, and, furthermore, the relationship between non-linear coded values and linear coded values is not as clearly systemized as in the case of the A-law, apart from the lowest value segment of the two curve halves.
If, in order to take into account the absence of systemization, the conversion from one form of coding into the other were approached along the lines corresponding to the A-law of treating PCM words located in the first segment of one-half of the curve in the same manner as PCM words located in the other segments, this would constitute a relatively unstructured system which could be achieved only with a relatively high technical expense.